:Rather than mess with a Popsicle (router on a stick) why not go with the
:2948-L3 (WS-C2948G-L3)? This switch seems to be one of Cisco's best kept
:secrets, but is in fact a very power and relatively inexpensive L3 switching
:solution. This switch has 48 FastEthernet ports and two 1000BaseX GBIC
:(Fiber) ports! There are only two downsides to this option. The first
:downside is that it does not support the integrated IOS. The integrated IOS
:combines the routing and switching IOS into one entity eliminating the need
:to switch between the SUP and Router module.
The 2948G-L3 only has a single accessible module. It uses the integrated
IOS. It can be a bit of a nuisance sometimes, as there are no commands
to configure a range of interfaces for a feature -- you have to go
into each interface and configure it separately.
: I've come to love the
:integrated IOS. The other downside is that if this sounds good you had
:better get one (two, three, the rest) now, while you still can because they
:have been targeted for EOL. ;-(
The link you provided said that the WS-C2926GS and WS-C2926GL are EOL'd,
not the WS-C2948G-L3. The only thing about the 2948G-L3 that
is EOL'd is the DC Power version: "Because of the unavailability of DC power
supplies, new orders cannot be accepted". (Same thing happened to
a number of different devices.)
Anyhow, the C2948G-L3 is interesting to be sure, but it does have
a few other weaknesses:
- blocks of 4 FE ports are controlled by the same ASIC, so there are some
things (e.g., encapsulation) that must be the same for each port
in a block
- no ACL's for FE ports. With newer releases, you can force all FE
traffic to be pushed up to the Gigabit port circuitry, to be CPU-
processed there by the same ACL controlling the incoming gigabit
traffic. This can only work properly if each FE port is part of
a different subnet, I believe. This feature is advertised as being
intended to prevent the FE ports from talking to each other.
- Some people have reported very port switching performance between
the FE ports, but good performance for routing. When I was testing
one, I did see very poor switching performance myself, but I did not
have time to chase down all the possibilities. If there is a way to
get fast switching performance, it certainly isn't *obvious*.
Some people have posted saying they now only recommend the 2948G-L3
for pure routing configurations.